reliability analysis of different rcied activation signal
Post on 24-Dec-2021
3 Views
Preview:
TRANSCRIPT
FACTA UNIVERSITATIS
Series: Electronics and Energetics Vol. 33, No 3, September 2020, pp. 459-476
https://doi.org/10.2298/FUEE2003459M
© 2020 by University of Niš, Serbia | Creative Commons License: CC BY-NC-ND
RELIABILITY ANALYSIS OF DIFFERENT RCIED ACTIVATION
SIGNAL RESPONSIVE JAMMING TECHNIQUES
AND THEIR COMPARISON TO ACTIVE JAMMING*
Mladen Mileusnić, Predrag Petrović, Vladimir Kosjer,
Aleksandar Lebl, Branislav Pavić
IRITEL a.d., Belgrade, Serbia
Abstract. In this paper we compared the time required for the successful jamming of
remote controlled improvised explosive devices activation using active and responsive
jamming methods. As a representative of active jamming method we analyzed jamming
signal generation using frequency sweep. For the analysis of the possible activating
signal presence based on responsive jamming procedures we first supposed Fast
Fourier Transform (FFT) implementation and compared its analysis rate to the rate of
sweep jamming. Taking into account the current technology state, it is proved that the
time required to achieve the successful jamming relied on FFT analysis may be less
than in the case of active sweep jamming. After that we considered pros and cons for
energy detector and matched filter detector implementation in responsive jamming. For
these two detector types it is shown how to determine the number of analysis blocks to
achieve approximately the same number of collected samples as in the case of FFT
implementation, starting from the probabilities of false detection and miss of detection.
Key words: Active and Responsive jamming, RCIED - remote controlled improvised
explosive devices, Frequency sweep, Fast Fourier Transform, Energy
Detector, Matched Filter, Jamming Reliability
1. INTRODUCTION
The common characteristic of all remote controlled improvised explosive devices
(RCIED) is that they are activated by wirelessly transmitted messages. The results of RCIED
activation message could be disastrous regarding people lives (VIP persons) and the
equipments damages. All elements related to activation signal characteristics (signal power,
frequency, implemented modulation method, message duration) are completely unknown.
This fact produces great problems in the realization of RCIED activation jammers.
Received February 4, 2020; received in revised form March 3, 2020
Corresponding author: Aleksandar Lebl
IRITEL a.d., 11080 Belgrade, Batajnički put 23, Serbia
E-mail: lebl@iritel.com *
The earlier version of this paper is awarded as the best one in the section Telecommunications at the 6th IcETRAN
Conference, Silver Lake, 3-6 June 2019, [1].
brought to you by COREView metadata, citation and similar papers at core.ac.uk
460 M. MILEUSNIĆ, P. PETROVIĆ, V. KOSJER, A. LEBL, B. PAVIĆ
Contributions [2] and [3] provide a general overview of jammers types, communications
jamming requirements and their efficiency analysis. Modern communications jamming
principles and techniques may be found in [4].
There are two basic approaches to the jammer implementation. The first one is active
jamming, which consisted of continuous predefined jamming signals sending independently
of the RCIED activating message characteristics. In this concept there are no „look through“
phases to detect the activation message existence and the jamming signal characteristics are
selected in general using previous experience and expectations. The most important freely
selected jamming signal parameter is the RF signal level. This level has to be as high as
possible to successfully prevent activating message reception. Two key features which are
not optimally chosen relate to continuous jamming regardless of RCIED activation message
existence and the RF jamming signal level necessary for jamming successfulness due to the
fact that the activation signal level is unknown.
The alternative approach to jammer implementation is responsive jamming concept.
In this case the jamming signal characteristics can be optimized using look through
intervals to detect the activation message existence and its level. That’s why it is possible
to send the jamming signal only during activation message presence and jamming signal
level can be adjusted to the activation message level in order to successfully deny the
threat. A wide range of active and responsive jammers may be found in [5]-[14].
It may be concluded from this short presentation of active and responsive jamming
characteristics that active jamming is always successful, while responsive jamming
efficiency depends on activation message detection reliability. The question is whether
responsive jamming reliability may be higher than for active jamming. In this paper we
compare the reliability of mostly implemented active jamming method – frequency
sweep [15]-[19] to the reliability of a representative method for activating signal eventual
presence detection in order to generate jamming signal according to the activation signal
characteristics by implementation of Fast Fourier Transform (FFT) in the analysis [14].
A brief principle schematic of RCIED activation signal detection is explained in Section
2. After that the method for RCIED activation signal frequency spectrum estimation based
on FFT analysis is presented in Section 3 with the emphasis on the required time for
calculation. Sections 4 and 5 deal with the specificities of energy detector and matched filter
detector implementation for RCIED activation message detection. The emphasis is on the
determination of collected samples number. Section 6 is devoted to frequency sweep
jamming and to determination of required time to realize one complete jamming cycle. In
Section 7 jamming reliability on the basis of FFT analysis is compared to the frequency
sweep jamming reliability, whereby two special purpose processors are considered for FFT
calculation. Reliability estimation is based on the required time to allow successful
jamming. Section 8 is focused on the presentation how to determine the necessary number
of analysis blocks in energy detection or matched filter detection to achieve the comparable
sample collection rate of these two detectors to FFT based detector. At the end, the paper
conclusion is given in Section 9.
2. PRINCIPLES OF DETECTION PROCESS
Main principles of RCIED activation signal detection may be explained using
simplified block-schema presented in Fig. 1.
Reliability Analysis of Different RCIED Activation Signal Responsive Jamming Techniques... 461
The first phase in detector function is signal samples collecting (block SCOL). After that
follows processing of these samples (block PROC). The final step is making a decision
about (eventual) presence of RCIED activation signal on the base of a set of comparison
rules (block DECISION). These comparison rules are adjusted to the applied method of
signal samples processing.
This paper is mainly devoted to the block PROC. The analyzed methods are FFT,
energy detector and matched filter. When the second or the third of these three methods
is implemented, digital filter precedes the phase of processing.
According to the available literature, there are also other methods which are less often
applied for spectrum sensing, but they are possible candidates for RCIED activation signal
detection. Some of them are waveform based detection, eigen-value based detection,
wavelet based edge detection, ciclostationary feature detection [20] and so on. These
methods, as generally less often applied ones, are beyond the scope of this paper.
SCOL PROC DECISION
Fig. 1 Block schema of RCIED activation signal detector
3. SIGNAL SPECTRUM ESTIMATION ON THE FFT BASE
FFT is the calculation procedure, which allows relatively fast estimation of discretized
signal frequency spectrum. Starting from n time samples of analyzed signal, this procedure
gives a snapshot of signal frequency spectrum also in n points, i.e. n spectrum lines are
obtained. FFT is the optimum method taking into account the required number of
mathematical operations for signal spectrum determination. There are (n/2)·log2 (n)
complex multiplications and n·log2 (n) complex additions [21]. The limitation for n is that
the condition n=2a must be satisfied, where a is the positive integer number. This is a
significant saving in the number of mathematical operations and in the required calculation
time comparing to the classical method of frequency spectrum estimation by Discrete
Fourier Transform (DFT). Namely, it is necessary to perform n2 complex multiplications
and n2–n complex additions to obtain n frequency spectrum components by DFT on the
base of n time samples.
Let us suppose that fs is the frequency of analyzed signal sampling. The sample
acquisition time is then:
s
ntf
(1)
The frequency resolution on the base of sample acquisition time may be determined as:
1 sfdfT n
(2)
Therefore, frequency resolution is improved when acquisition time is increased, i.e. the
space between frequency spectral components of the analyzed signal is lower.
462 M. MILEUSNIĆ, P. PETROVIĆ, V. KOSJER, A. LEBL, B. PAVIĆ
Constant advancements in processor realization technology and mathematical algorithm
improvements are visible in two aspects of FFT calculation progress. On the one side the
number of points in which frequency spectrum is determined is constantly increased, and on
the other side the required FFT calculation time for some exactly determined number of
frequency spectrum components is constantly decreased, chronologically, successively
according to presentations in [22], [23], [21], [24]. We selected two approaches referenced
in [21] and [24] due to very fast processing algorithms.
Data presented in [24] is related to the FFT calculation time as a function of the
number of signal time samples implemented for FFT calculation, i.e. as a function of the
obtained frequency components number in the analyzed spectrum. The presented data is
for processor clock of 1GHz. It is further emphasized in [24] that improvement may be
achieved by processor clock speed increase to 1.25GHz. Besides, it is stated in [25] that
maximum processor clock frequency may be even 1.4GHz. On the base of these data, the
FFT calculation time (Tcal in ms) is presented in Table 1 as a function of the number of
points used in a calculation, for a processor clock of 1.25GHz and for 8 processor cores.
The value of the constant K is 1024 in the first column of the Table 1.
The time of FFT calculation (Tcal in μs) according to the data emphasized in [21] is
presented in the Table 2. The processor clock in this case may be in the range between
60MHz and 150MHz [26]. That’s why data are presented for the mean processor
frequency of 100MHz. FFT hardware accelerator (HWAFFT) is one of the parts in the
processor implemented according to [21]. HWAFFT is intended for faster FFT
calculation. Data in Table 2 are related to the case when HWAFFT is implemented. The
number of points is relatively small (till 1024) where FFT is calculated comparing to the
number of points, where FFT results are presented in Table 1.
In accordance to Fig. 1, the total time, which is needed for signal analysis in a jammer
(Tan) before (eventually) starting RCIED activation jamming signal emission, consists of
three components: sample acquisition time (T), FFT calculation time (Tcal) and the time,
which is necessary to compare obtained signal frequency components after FFT
calculation (Tcomp) in order to determine whether it is necessary to start jamming. When
considering the last component (Tcomp), there is not such a data in a literature, because
calculation is very specific. For our analysis, we supposed that taking equal values of Tcal
and Tcomp is a quite good approximation, i.e.
2an cal comp cals
nT T T T Tf
(3)
Table 1 The time of FFT calculation as a function of the number of calculation points for
the processor presented in [24]
Number of points for FFT calculation Calculation time Tcal [ms] (8 cores, 1.25GHz)
16K 0.1051
32K 0.1584
64K 0.2517
128K 0.5128
256K 0.9488
512K 2.4824
1024K 5.1226
Reliability Analysis of Different RCIED Activation Signal Responsive Jamming Techniques... 463
Table 2 The time of FFT calculation as a function of the number of calculation points for
the processor presented in [21]
Number of points for FFT calculation Calculation time Tcal [μs] (with HWAFFT, 100 MHz)
8 1.3
16 1.7
32 3.21
64 4.36
128 9.12
256 16.68
512 37.4
1024 73.15
4. RCIED ACTIVATION SIGNAL DETECTION BY ENERGY DETECTOR
Energy detector is the simplest techniques for signal detection [27]. In the same time
it is a very often applied technique. It is necessary first to measure signal energy in the
pre-defined frequency band. The measured signal energy is then compared to the energy
threshold according to the equation
2
1
( ) ( ( ))
N
n
E x x n
(4)
where N is the number of samples implemented for signal energy estimation, x(n) is the
amplitude of nth sample and γ is the threshold power.
Although simple for implementation, energy detector performances are degraded due to
noise uncertainty (noise level is variable during time) and background interference [28].
Noise uncertainty may be bounded or unbounded [29]. As a consequence of noise
uncertainty, the detection by energy detector may become even impossible under relatively
low value of signal to noise ratio (SNR) [30]. In other words, there exists a SNR wall:
detection is possible only when signal power is higher than noise power uncertainty.
For the analysis in this paper and for the comparison of energy detector characteristics
with the characteristics of other methods for reactive jamming the most important parameter
is the number of samples (N) to achieve necessary detection reliability. Our analysis is based
on the formula for N from [27] [31]:
1 1 2
2
( ( ) 1 2 ( ))2
f dQ P SNR Q PN
SNR
(5)
where Pf is probability of false detection (detector announces signal presence although
there is no signal), Pd is probability of successful detection and Q-1 is inverse Gaussian-Q
function. In other words, Q-1 is the inverse of
2
121( )
2
u
x
Q x e du
(6)
464 M. MILEUSNIĆ, P. PETROVIĆ, V. KOSJER, A. LEBL, B. PAVIĆ
1,E+01
1,E+02
1,E+03
1,E+04
1,E+05
1,E+06
1,E+07
-20 -19 -18 -17 -16 -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0
SNR(dB)
N
Pf=Pmd=0.1 Pf=Pmd=0.01 Pf=Pmd=0.001 Pf=Pmd=0.000001
Fig. 2 The necessary number of samples (N) when energy detector is applied as a function
of signal-to-noise (SNR) ratio without noise uncertainty for different values Pmd =Pf.
One additional important parameter in energy detection systems analysis is miss in
detection Pmd (detector does not detect a signal although it exists). Probabilities Pd and
Pmd are connected by the equation
1md dP P (7)
Fig. 2 presents the necessary number of samples (N) as a function of signal-to-noise ratio
(SNR). The results are presented for equal values of Pf and Pmd. There is no noise uncertainty
which means that optimum detector threshold value exists independently of SNR. For small
SNR signal detection is always possible, but the value of N significantly increases.
In our concrete implementation it is more important to achieve low value of Pmd than
to achieve low value of Pf. In other words, consequences of miss in detection are more
severe (RCIED is activated because there is no jamming) than if the detection is false
(only jamming signal is waste generated). That’s why the results for probability values
satisfying the condition Pmd<Pf are presented in Fig. 3.
1,E+01
1,E+02
1,E+03
1,E+04
1,E+05
1,E+06
1,E+07
-20 -19 -18 -17 -16 -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0
SNR(dB)
N
Pf=0.1;Pmd=0.01 Pf=0.01;Pmd=0.001 Pf=0.001;Pmd=0.000001 Pf=0.1;Pmd=0.01;ro=0.25dB
Fig. 3 The necessary number of samples (N) when energy detector is applied as a function of
signal-to-noise (SNR) ratio without noise uncertainty for different values Pmd <Pf.
Reliability Analysis of Different RCIED Activation Signal Responsive Jamming Techniques... 465
Noise uncertainty is modelled in such a way that the value ρ-(1/ρ), where it is ρ>1, is
subtracted from the value SNR in the denominator of equation (5). This means that noise
power, instead of having power equal to σ2 when noise is completely defined, now has
the value of power between (1/ρ)·σ2 and ρ·σ2. SNR-wall is presented by the fourth graph
in Fig. 3. Even for a very small noise uncertainty value ρ=0.25dB or ρ=1.059 when it is
Pf=0.1 and Pmd=0.01 the value of N tends to infinity for SNR~-9.3dB and below this
value -9.3dB it is not possible to detect a signal. As a conclusion it may be said that it is
very important to constantly monitor the noise level and to adjust threshold value
according to instantaneous noise level and in this way to avoid SNR-wall appearance.
5. RCIED ACTIVATION SIGNAL DETECTION BY MATCHED FILTER
The second often implemented technique of spectrum analysis is based on the method of
matched filters. The main property of such filters is that they are optimum linear filters
applied for signal detection in white Gaussian noise, meaning that maximum SNR is
achieved by their implementation [32]. Although this property contributes to easier and
faster signal detection, the drawback of matched filter implementation is that it is necessary
to precisely know time characteristics of the signal which has to be detected. Such
knowledge is possible in some implementation areas, as for example, in cognitive radio
[32], [33]. But, if considering RCIED activation signal jamming, there is a great variety of
possible and, in the same time, unpredictable activation techniques. They usually depend on
the devices which may be easily purchased in some country (region) and easily adapted for
its malicious function. The number of applied solution types is not great in the analysis
presented in [34] with the dominant implementation of one type, thus simplifying and
limiting the necessary number of different matched filters. Nevertheless, application of
matched filters is not quite suitable for RCIED activation signal detection and the analysis
of this method has more theoretical than practical significance.
Similarly to the analysis of energy detector, the necessary number of samples to
achieve the desired probability of false alarm and probability of successful signal
detection may be determined on the base of equation from [27]:
1 1 2( ( ) ( ))f dQ P Q PN
SNR
(8)
Fig. 4 presents the necessary number of samples (N) as a function of signal-to-noise
ratio (SNR) when matched filter is implemented. The results are obtained by equation (8)
and are presented for equal values of Pf and Pmd. After that, Fig. 5 presents the
corresponding results if Pmd<Pf. It is necessary to collect and analyze lower number of
samples to achieve the same values Pf and Pmd as when energy detector is implemented.
This is noticeable if graphs from Figs. 2 and 4, as well as graphs from Figs. 3 and 5 are
mutually compared.
466 M. MILEUSNIĆ, P. PETROVIĆ, V. KOSJER, A. LEBL, B. PAVIĆ
1,E+00
1,E+01
1,E+02
1,E+03
1,E+04
-20 -19 -18 -17 -16 -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0
SNR(dB)
N
Pf=Pmd=0.1 Pf=Pmd=0.01 Pf=Pmd=0.001 Pf=Pmd=0.000001 Fig. 4 The necessary number of samples (N) when matched filter is applied as a function
of signal-to-noise (SNR) ratio for different values Pmd =Pf.
1,E+01
1,E+02
1,E+03
1,E+04
-20 -19 -18 -17 -16 -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0
SNR(dB)
N
Pf=0.1;Pmd=0.01 Pf=0.01;Pmd=0.001 Pf=0.001;Pmd=0.000001
Fig. 5 The necessary number of samples (N) when matched filter is applied as a function
of signal-to-noise (SNR) for different values Pmd <Pf.
6. ACTIVE JAMMING USING FREQUENCY SWEEP
Frequency sweep is often used method of active jamming. It is necessary to linearly
change signal frequency step by step from its minimum value (f1) to the maximum one (f2)
in order to realize a sweep. It is a readily implemented jamming method, because a
significantly smaller power is necessary in relation to wideband jamming based on Additive
White Gaussian Noise (AWGN) [17] - [19]. Linear frequency change of jamming signal
frequency is practically approximated by a stepwise change, as it is presented in Fig. 6.
There are two parameters, besides outmost sweep signal frequencies f1 and f2, which model
signal frequency change: frequency change step (f∆) and each step time interval duration
while the same signal frequency is generated (T∆).
Reliability Analysis of Different RCIED Activation Signal Responsive Jamming Techniques... 467
f
f1
f2
t
τsw
Wsw
f∆
T∆
Fig. 6 Practical implementation of frequency sweep in the RCIED activation message
active jamming.
The total step number in jamming realization may be represented by an equation:
2 1f fN
f
(9)
while the duration of one total sweep cycle may be represented as:
2 1sw
f fT N T T
f
(10)
7. COMPARISON OF THE ACTIVE AND RESPONSIVE FFT JAMMING RELIABILITY
The starting data in our analysis will be the required time to realize sweep jamming of
RCIED activation signal under the condition that at least once jamming signal frequency
and RCIED activation message frequency are approximately equal. After that we shall
determine the total time from the beginning of the analyzed signal sample acquisition
including FFT calculation time until the start of jamming signal emission on the detected
RCIED activation signal frequency. In order to achieve comparison requirements of these
two results, we shall suppose that the number of steps in sweep procedure (N) is equal to
the number of points where the analyzed signal spectrum is estimated (n).
There is a number of D/A converters, which may be implemented for jamming signal
generation. One typical example may be found in [35]. This D/A converter is used in our
jammer solution [16]. It generates analog signal from the samples, whose maximum
frequency is pretty high (fs=3.5GHz), thus enabling maximum generated signal frequency
fg=1.4GHz.
Let us suppose that RCIED activation jamming is realized by jamming signal generation
in a frequency band between f1=20MHz and f2≈1.33GHz. We shall further define the
468 M. MILEUSNIĆ, P. PETROVIĆ, V. KOSJER, A. LEBL, B. PAVIĆ
frequency change step f∆=20kHz, and the frequency sweep step duration T∆=200ns (very
short time, a greater time is used in a practical realization - T∆=1μs or more [16]). On the
base of (9) we obtain the number of steps in a sweep procedure realization N=65536=64K,
and on the base of (10) sweep procedure duration is Tsw≈13.1ms.
Let us now have a device for responsive jamming, which collects the analyzed signal
samples (performs A/D conversion) at the same frequency fs=3.5GHz as the frequency of
samples for D/A converter [36]. Number of samples that need to be collected, and
consequently obtained number of the analyzed signal discrete frequency components is the
same as the number of sweep steps (n=65536=64K). In this way it is achieved that the
accuracy of jamming signal frequency df according to (2) is the same as the step of sweep
signal frequency change f∆ according to (9). In the case of responsive jamming, the signal
sample acquisition time from (1) will be T≈18.7μs, while on the basis of Table 1 the FFT
calculation time is Tcal=251.7μs. According to our adopted approximation it is also
Tcomp=251.7μs. Taking into account these three values, the total analysis time on the basis of
(3) is Tan=522.1μs.
Comparing two time intervals which are the main indicators of active and responsive
jamming reliability (Tsw and Tan), it is concluded that responsive jamming on the basis of
FFT implementation is significantly more reliable (in our example ≈25 times) than active
jamming by frequency sweep. One additional element which is an advantage of responsive
jamming is the fact that after the analysis process (Tan) jamming may be performed only on
the detected activation signal frequency with no time limit. In the case of active jamming,
signal frequency is only during a short time period T∆ approximately equal to the activation
signal frequency. Therefore, due to greater jamming time it is also higher the probability
that jamming is successful when responsive jamming is implemented.
Fig. 7 presents the results of reliability comparison of responsive jamming based on
FFT analysis according to data from [24] in relation to active jamming based on
frequency sweep. When active jamming is considered, jamming time on each frequency
is adopted to be 200ns. Two extreme cases are chosen for responsive jamming: a) the
first one, which allows maximum analysis speed (maximum specified processor clock
0
5
10
15
20
25
30
35
16K 32K 64K 128K 256K 512K 1024K
N FFT
Ts
w/T
an
.
1.4GHz 8 cores 1GHz 1 core
Fig. 7 The reliability of responsive jamming on the FFT basis according to [24] in relation
to active frequency sweep jamming.
Reliability Analysis of Different RCIED Activation Signal Responsive Jamming Techniques... 469
frequency 1.4GHz and maximum number of processor cores – 8) and b) the second one,
which corresponds to the minimum analysis speed (minimum processor clock frequency
1GHz and minimum number of processor cores – 1). The required number of processor
cycles for FFT calculation as a function of the number of FFT points is taken according
to Table 1 from [24]. After that, it is determined the necessary time for FFT analysis in
some cases. The graph in Fig. 7 indicates higher reliability of responsive jamming on the
base of FFT in relation to active jamming using frequency sweep in all jamming
conditions, because the calculated relation of jamming reliability is always between two
extreme cases presented in Fig. 7 (i.e. this relation is always greater than 1).
Fig. 8 presents the results of reliability comparison of responsive jamming based on
FFT analysis according to data from [21] in relation to active jamming based on
frequency sweep. The processor implemented for FFT calculation operates on lower
frequencies (between 60MHz and 150MHz) [26] than in the example from Fig. 7. That’s
why it is adopted that analyzed signal sampling is realized on a significantly lower
frequency (800MHz) than in the example in Fig. 7. This further means that jamming is
realized in this case for lower frequencies (till 320MHz).
0
0,5
1
1,5
2
2,5
8 16 32 64 128 256 512 1024N FFT
Tsw
/Tan
.
150MHz with HWAFFT 60MHz no HWAFFT 100MHz with HWAFFT 100MHz no HWAFFT
Fig. 8 The reliability of responsive jamming on the FFT basis according to [21] in
relation to active frequency sweep jamming.
The results in Fig. 8 are presented separately in the case that HWAFFT is used for an
analysis and when it is avoided. The maximum processing rate in this case is achieved if
HWAFFT is used together with maximum processor clock frequency (150MHz), while the
minimum processing rate is if HWAFFT is not used and the processor clock frequency is
minimum (60MHz). These results are presented by first two vertical graphs for each number
of frequencies in FFT analysis. Besides these two graphs, the results for mean processor
clock frequency (100MHz) are presented when HWAFFT is used and when it is not used.
The required number of processor cycles to calculate FFT for some number of points in
FFT analysis is determined on the base of Table 3, i.e. Table 4 from [21].
Using the analysis the results in Fig. 8 it can be concluded that the application of
HWAFFT allows also in this case that responsive jamming using FFT may be more
reliable than the active jamming by frequency sweep (except for the smallest number of
470 M. MILEUSNIĆ, P. PETROVIĆ, V. KOSJER, A. LEBL, B. PAVIĆ
analyzed frequencies – 8, which is unlikely to occur in practice). On the contrary, if
HWAFFT is not used, responsive jamming reliability using FFT becomes lower than the
reliability of frequency sweep jamming (because the relation Tsw/Tan<1). For the mean
processor clock frequency (100MHz) and with the HWAFFT implementation for the
smaller number of points in the analysis, frequency sweep implementation is more
reliable, while for the greater number of points the analysis on the base of FFT is better.
8. PERFORMANCES OF ENERGY DETECTOR AND MATCHED FILTER DETECTOR
ON THE BASE OF MAIN FFT PARAMETERS
In our further analysis we are going to investigate achievable performances of energy
detector and matched filter detector on the base of 64K collected (analyzed) signal
samples, as in the case of FFT analysis. In this way analysis procedure duration is
comparable for these two methods. The time of sample collection is exactly the same as
we suppose the same sampling rate in two presented cases.
Fig. 9 presents the simplified timing diagram of a jammer realized on the base of energy
detector. The value SNR is relatively low when the whole available frequency bandwidth is
analyzed at once, such that the necessary number of samples (N) for the analysis is higher
than 64K. That’s why the complete frequency bandwidth of the analyzed system is separated
into n distinguished blocks, meaning that bandpass filtering (BPF) is the first step at each of n
executed block inputs for energy detection (ED). At the output of each block is comparator
(COMP) and after that (in the block DECISION) is determined whether RCIED activation
signal is present. The principle structure is the same if energy detector is replaced by matched
filter (MF), except that designations ED1...EDn in Fig. 9 have to be replaced by MF1...MFn.
There are two algorithm execution possibilities: 1. sequential processing in each of n blocks
(Fig. 9a)); 2. parallel processing in n blocks when considering the elapsed time (Fig. 9b)). Of
course, the combination of these two structures is also possible, but this time scenario is
beyond the scope of our analysis. If the white noise power in the whole available frequency
bandwidth is designated by σv2, the power at the output of each of n bandpass filters is σv
2/n,
because white noise frequency spectrum is considered to be uniform and it is split into n
equal frequency portions. It means that SNR in the energy detector where RCIED activation
signal appears is n times higher than without spectrum separation.
BPFnt
EDn COMPn DECISIONn BPF1 ED1 COMP1 DECISION1...
a)
BPFn
t
EDn COMPn DECISIONn
BPF1 ED1 COMP1 DECISION1
...
...
...
...
b)
Fig. 9 The timing diagram of the energy detector based jammer: a) sequential processing;
b) parallel processing
Reliability Analysis of Different RCIED Activation Signal Responsive Jamming Techniques... 471
Fig. 10 presents the total number of analyzed samples (N) to successively perform
energy detection as a function of the number of blocks (frequency spectrum parts) n when
SNR=-20dB for the whole frequency bandwidth in the case that sequential processing is
applied as in Fig. 9a). The total number of collected samples is determined as a product of
the number of sequentially realized blocks and the necessary number of samples for each
block to achieve the noise power σv2/n. The results are presented for three pairs of values Pf
and Pmd when there is no noise power uncertainty and for one of these parameter pairs when
the value of noise power uncertainty is ρ=0.25dB. We want to have 64K analyzed samples
as in the case of FFT based detection and we may determine the necessary number of
sequential blocks for different Pf and Pmd. For example, when it is Pf=0.1 and Pmd=0.01, it is
enough to have n=5 sequential blocks when there is no noise power uncertainty. This value
increases to even n=24 when noise power uncertainty is only ρ=0.25dB. It is expected that
such a low ρ is very rare, because there are usually a lot of different signals besides white
noise in the wireless system surrounding.
1,E+03
1,E+04
1,E+05
1,E+06
1,E+07
1 10 100
number of sequential blocks
N
Pf=0.1;Pmd=0.01 Pf=0.01;Pmd=0.001 Pf=0.001;Pmd=0.000001 Pf=0.1;Pmd=0.01;ro=0.25dB
Fig. 10 The necessary number of analyzed samples (N) as a function of the number of
analysis blocks in the case of energy detector implementation when SNR=-20dB
for the whole frequency bandwidth when sequential processing is used.
Fig. 11 presents the total number of analyzed samples (N) to perform energy detection as
a function of the number of blocks (frequency spectrum parts) n when SNR=-20dB for the
whole frequency bandwidth in the case that parallel processing is applied as in Fig. 9b). The
parameters on this figure are the same as for Fig. 10. If there is no noise power uncertainty, it
is enough to have only 2 parallel blocks and to collect about 64K samples when it is Pf=0.1
and Pmd=0.01. With the noise power uncertainty we ought to have 14 blocks.
472 M. MILEUSNIĆ, P. PETROVIĆ, V. KOSJER, A. LEBL, B. PAVIĆ
1,E+01
1,E+02
1,E+03
1,E+04
1,E+05
1,E+06
1,E+07
1 10 100
number of parallel blocks
N
Pf=0.1;Pmd=0.01 Pf=0.01;Pmd=0.001 Pf=0.001;Pmd=0.000001 Pf=0.1;Pmd=0.01;ro=0.25dB
Fig. 11 The necessary number of analyzed samples (N) as a function of the number of
analysis blocks in the case of energy detector implementation when SNR=-20dB
for the whole frequency bandwidth when parallel processing is used.
Fig. 12 presents the total number of analyzed samples (N) to perform matched filter
detection as a function of the number of blocks (frequency spectrum parts) n when SNR=-
40dB for the whole frequency bandwidth in the case that sequential processing is applied as
in Fig. 9a). The conditions for the graphical presentation are the same as in Fig. 10. In this
case the number of necessary points for the analysis is independent of the number of applied
sequential blocks. Such behaviour is the consequence of the fact that SNR is linear factor in
the denominator of equation (8). It means that algorithm characteristics may not be
improved by sequential processing for matched filter implementation. Opposite to this, SNR
appears as the quadratic factor in the denominator of equation (5) for energy detector. Also
SNR appears under the square root of the numerator in the same equation. Such complex
dependence causes decreasing of the number of necessary samples when the number of
analysis blocks increase in the case of energy detector implementation.
1,E+05
1,E+06
1 10 100
number of sequential blocks
N
Pf=0.1;Pmd=0.01 Pf=0.01;Pmd=0.001 Pf=0.001;Pmd=0.000001 Fig. 12 The necessary number of analyzed samples (N) as a function of the number of
analysis blocks in the case of matched filter detector implementation when SNR=
-40dB for the whole frequency bandwidth when sequential processing is used.
Reliability Analysis of Different RCIED Activation Signal Responsive Jamming Techniques... 473
Fig. 13 presents the total number of analyzed samples (N) to perform matched filter
detection as a function of the number of blocks (frequency spectrum parts) n when SNR=-
40dB for the whole frequency bandwidth in the case that parallel processing is applied as in
Fig. 9b). The results are presented for hundred times worse value of SNR than in the case of
energy detector consideration. This is one more proof of the better matched filter
performances comparing to energy detector.
1,E+03
1,E+04
1,E+05
1,E+06
1 10 100
number of parallel blocks
N
Pf=0.1;Pmd=0.01 Pf=0.01;Pmd=0.001 Pf=0.001;Pmd=0.000001
Fig. 13 The necessary number of analyzed samples (N) as a function of the number of
analysis blocks in the case of matched filter detector implementation when SNR=
-40dB for the whole frequency bandwidth when parallel processing is used.
Energy detector or matched filter detector as the base of RCIED activation jammer are
more suitable in the case when there is necessary to intercept only a frequency spectrum
part. In such a case the number of sequential or parallel blocks on the block-schema from
Fig. 9 is decreased as a consequence of increased SNR, making a solution practically
realisable. This situation is presented in [34], where it is demonstrated that even in the area
of significant military war activities, there is not too high number of implemented RCIED
activation solution types. Therefore, if energy detector or matched filter detector supervises
the wideband signal, the number of spectrum parts to which the whole frequency bandwidth
is separated is relatively low (not more than 100 according to figures 10-13). As a
consequence, each frequency part for detection is relatively wide and it may be efficiently
jammed only by frequency limited noise (barrage) jamming. Opposite, FFT gives a lot of
points where frequency spectrum is estimated (64K in the example in Section 7). That’s
why jamming may be realized on one or more precisely defined frequencies implementing
pure sinusoid (spot) jamming.
The pipeline processing is one possibility to several times increase the analysis rate
without significantly increasing hardware size. This possibility may be considered first of
all when FFT method is applied, but also when energy detector or matched filter detector
are considered. The principle details about pipeline processing in RCIED detection are
found in [1], [37].
474 M. MILEUSNIĆ, P. PETROVIĆ, V. KOSJER, A. LEBL, B. PAVIĆ
8. CONCLUSIONS
The results of the calculation presented in this paper have proved that responsive
jamming may be more reliable than active jamming. This is the first paper contribution. The
required time for secure jamming signal generation on the frequency of RCIED activation
message is analyzed as a criterion of jamming reliability. As the result of complete analysis
based on FFT, the frequency of RCIED activation signal is obtained, and jamming on the
exactly determined frequency may be initiated. The results are compared for two
processors, which are specialized for FFT calculation. One of these two processors provide
that responsive jamming based on FFT implementation is more reliable than active jamming
using frequency sweep. In this case analysis rate is several times, and even up to several
tens of times higher when FFT is implemented in the analysis in relation to the frequency
sweep rate. For the second analyzed processor reliability of responsive jamming depends on
processor hardware characteristics such as the processor clock frequency and whether
hardware accelerator (HWAFFT) is applied. If HWAFFT is included in the analysis with
the higher processor clock frequency, jamming based on FFT analysis is certainly more
reliable. On the contrary, if HWAFFT is not used with the lower processor clock frequency,
the speed of FFT analysis may not approach jamming speed realized by frequency sweep.
The second paper contribution is related to the possibilities of energy detector and
matched filter detector implementation for RCIED activation responsive jamming. It is
proved that these jamming types are more suitable for jamming of narrower frequency
bands, because frequency spectrum is estimated in significantly lower number of points
than it is the case with FFT analysis. The additional problem for energy detector and
matched filter detector implementation is a priori unknown shape of RCIED activation
signal and noise power level. To overcome these problems, these two detectors require
collecting relatively high number of signal samples. It is explained how this number of
samples may be reduced while considering the desired probabilities of false detection and
miss of detection, as well as analysis rate (number of collected samples).
It can be summarized that the results of comparative analysis presented in this paper
prove that at up-to-date technological development level, RCIED activation responsive
jamming may be very reliable and very often even more reliable than active jamming,
especially when FFT analysis is considered.
This paper is the extended and enhanced version of the contribution [1]. Comparing
to [1], completely new sections are 2, 4, 5 and 8. Section 2 presents the principle block-
schema of the analyzed solutions to simplify readers following the text as two new
methods are described in the paper. These two new methods are introduced in Sections 4
and 5. The priority in the analysis is given to the determination of necessary number of
samples to achieve comparable timing performances to the RCIED activation signal
detector realized by FFT. The main new results for practical jammer realization are in the
Section 8, where it is analyzed how to choose the most important jammer specification
parameters: the necessary number of collected samples and the number of sequential or
parallel blocks to achieve the desired jammer characteristics.
Acknowledgement: The paper is realized in the framework of the project TR32051, which is
cofinanced by Ministry of Education, Science and Technological Development of the Republic of Serbia,
2011-2019.
Reliability Analysis of Different RCIED Activation Signal Responsive Jamming Techniques... 475
REFERENCES
[1] M. Mileusnić, P. Petrović, A. Lebl, B. Pavić, “Comparison of RCIED Activation Responsive and Active
Jamming Reliability”, In Proceedings of the 6th International Conference IcETRAN 2019, Silver Lake,
June 3-5, 2019, pp. 988–993.
[2] P. Petrović and M. Šunjevarić, “Radio Surveillance and Jamming Systems and Techniques“, IN
Proceedings of EEE Conference: Trends in telecommunications development (organized by Faculty of
Electrical Engineering and the company Electronics Industry), Belgrade, November 1988, pp. 17.1–
17.22., in Serbian.
[3] P. Petrović, “Syllabus EW Course: Introductory Course of Electronic Warfare Systems, Techniques and
Technologies”, Preliminary material for lectures at the Royal Academy in Shrivenham, Research Gate,
January 2011.
[4] R. Poisel, Modern Communications Jamming Principles and Techniques. Second Edition, Artech House,
Boston/London, 2011.
[5] Security and Defense Technologies (Secintel), “Portable RF Jammer“, http://www.secintel.com/ecom-
prodshow/portable_rf_jammer.html.
[6] Homeland Security Strategies GB LTD, “VIP-300T Covert IED Jammer – Trunk Mounted Version“,
http://www.secintel.com/media/pdf/vip300T_Covert_IED_Jammer.pdf.
[7] Security and Defense Technologies (Secintel), “Reseller Demonstration Kits“, http://www.secintel.com/ecom-
prodshow/stationary_jammer.html.
[8] Security and Defense Technologies (Secintel), “Backpack Jammer“, http://www.secintel.com/ecom-
prodshow/backpack_jammer.html.
[9] HSS Development, “Technologies + Solutions“, http://www.secintel.com/media/pdf/hsscatalog.pdf.
[10] Phantom Technologies LTD., “Our Security Solutions”, http://www.phantom.co.il/ufiles/attaches/
RCJ1390LT-H_new.pdf.
[11] Elisra Electronic Systems ltd, “EJAB Family Electronic Jammers Against Bombs“, http://www.mw-
elisra.com/pdf/EJAB-BrochureLIGHT001.pdf.
[12] Aselsan, “GERGEDAN – Portable RCIED Jammer System (Vehicle Type)“, http://www.aselsan.com.tr/en-
us/capabilities/electronic-warfare-systems/electronik-support-and-electronic-attack-systems/gergedan-portable-
rcied-jammer-system-%28vehicle-type%29.
[13] J. Mietzner, P. Nickel, A. Meusling, P. Loos and G. Bauch, “Responsive communications jamming
against radio-controlled improvised explosive devices”, IEEE Communications Magazine, vol. 50, no.
10, pp. 38–46, October 2012.
[14] K. Wilgucki. R. Urban, G. Baranowski, P. Grądzki, P. Grądzki and P. Skarźyński, “Automated protection
system against RCIED”, In Proc. Military Communication Institute (MCI), 2012.
[15] P. Petrović, N. Remenski, P. Jovanović, V. Tadić, B. Pavić, M. Mileusnić and B. Mišković, “WRJ 2004
Wideband Radio Jammer against RCIEDs”, technical solution – Project TR32051, 2011,
http://www.iritel.com/images/pdf/wrj2004-e.pdf.
[16] M. Mileusnić, P. Petrović, B. Pavić, V. Marinković-Nedelicki, J. Glišović, A. Lebl and I. Marjanović,
“The Radio Jammer Against Remote Controlled Improvised Explosive Devices“, In Proceedings of the
25th Telecommunications Forum (TELFOR), Belgrade, November 21-22th 2017, pp. 151–154.
[17] M. Mileusnić, B. Pavić, V. Marinković-Nedelicki, P. Petrović, D. Mitić and A. Lebl, “Analysis of
Jamming Successfulness against RCIED Activation“, In Proceedings of the 5th International Conference
IcETRAN 2018, Palić, June 11-14th 2018, pp. 1206–1211. Paper awarded as the best one in the section
Telecommunications.
[18] M. Mileusnić, P. Petrović, B. Pavić, V. Marinković-Nedelicki, V. Matić and A. Lebl, “Jamming of
MPSK Modulated Messages for RCIED Activation”, In Proceedings of the 8th International Scientific
Conference on Defensive Technologies OTEH 2018, Belgrade, 11-12th October 2018, pp. 380–385.
[19] M. Mileusnić, B. Pavić, V. Marinković-Nedelicki, P. Petrović, D. Mitić and A. Lebl, “Analysis of jamming
successfulness against RCIED activation with the emphasis on sweep jamming,“ Facta Universitatis, Series
Electronics and Energetics, Vol. 32, No. 2, June 2019, https://doi.org/10.2298/FUEE1902211M, ISSN: 0353-
3670, pp. 211-229., the extended and revised version of the paper from the IcETRAN 2018.
[20] R. R. Jaglan, S. Sarowa, R. Mustafa, S. Agrawal, N. Kumar, “Comparative Study of Single-user Spectrum
Sensing Techniques in Cognitive Radio Networks”, In Proceedigs of the Second International Symposium on
Computer Vision and the Internet (VisionNet’15), Procedia Computer Science 58, 2015, pp. 121–128.
[21] M. McKeown, “Texas Instruments: FFT Implementation on the TMS320VC5505, TMS320C5505, and
TMS320C5515 DSPs“, Application Report SPRABB, June 2010 – Revised January 2013, pp. 1-28.
476 M. MILEUSNIĆ, P. PETROVIĆ, V. KOSJER, A. LEBL, B. PAVIĆ
[22] E. Çetin, R. C. S. Morling and I. Kale, “An Integrated 256-point Complex FFT Processor for Real-time
Spectrum Analysis and Measurement”, In Proceedings of the IEEE Proceedings of Instrumentation and
Measurement Technology Conference, May 1997, vol. 1, pp. 96–101.
[23] “Pipelined FFT/IFFT 256 points (Fast Fourier Transform) IP Core User Manual“, Unicore Systems Ltd., pp. 1-
17., https://opencores.org/websvn/filedetails?repname=pipelined_fft_256&path=%2Fpipelined _fft_256%
2Ftrunk%2FDOC%2Ffft256_um.pdf.
[24] X. Li, E. Blinka, “Very large FFT for TMS320C6678 processors“, Texas Instruments, 2015., pp. 1-6.
[25] Texas Instruments, “Multicore Fixed and Floating-Point Digital Signal Processor“, SPRS691 –
November 2010 – Revised March 2014., pp. 1-242.
[26] Texas Instruments, “TMS320C5505 Fixed-Point Digital Signal Processor“, SPRS660F – August 2010 –
Revised September 2013., pp. 1-157.
[27] T. Trump and I. Müürsepp, “Detection Speed of Responsive Communication Jamming Detectors, Recent
Advances in Telecommunications and Circuits”, 2nd International Conference on Circuits, Systems,
Communications, Computers and Applications, Dubrovnik, June 2013, pp. 149-154.
[28] S. Atapattu, C. Tellambura and H. Jiang, Energy Detection for Spectrum Sensing in Cognitive Radio.
Springer, 2014, Chapters 1-2, ISSN 2191-5768, ISBN 978-1-4939-0493-8, DOI 10.1007/978-1-4939-
0494-5, pp. 1-26.
[29] S. Bahamou and A. Nafkha, “Noise Uncertainty Analysis of Energy Detector: Bounded and Unbounded
Approximation Relationship”, In Proceedings of the 21st European Processing Conference (EUSIPCO
2013), 9-13. September 2013, Marrakech, Morocco, pp. 1–4.
[30] R. Tandra and A. Sahai, “SNR walls for signal detection”, IEEE Journal on Selected Topics in Signal
Processing, vol. 2, no. 1, pp. 1–30, 2008.
[31] M. M. Tanatwy, “Responsive Communication Jamming Detector with Noise Power fluctuation using
Cognitive Radio”, International Journal of Innovative Research in Computer and Communication
Engineering, vol. 2, no. 10, pp. 5967–5973, October 2014.
[32] S. Dannana, B. Prasad Chapa, G. S. Rao, “Spectrum Sensing Using Matched Filter Detection”, Intelligent
Engineering Informatics: Proceedings of the 6th International Conference on Frontiers of Intelligent
Computing: Theory and Applications (FICTA), 2017, Bhubaneswar, Odisha, Springer Verlag, Singapore,
pp. 497-503, ISBN: 9811075654.
[33] F. Salahdine, H. E. Ghazi, N. Kaabouch, W. F. Fihri, “Matched Filter Detection with Dynamic Threshold
for Cognitive Radio Networks”, In Proceedings of the 2015 International Conference on Wireless
Networks and Mobile Communications (WINCOM), 23th October 2015, France.
[34] G. Attila, “The Radio Controlled Improvised Explosive Device (RCIED) threat in Afghanistan, AARMS,
vol. 12, no. 1, 2013, pp. 9–23.
[35] Analog Devices, “AD9914: 3.5 GSPS Direct Digital Synthesizer with 12-Bit DAC“, Rev. F, February
2017., pp. 1-46.
[36] Texas Instruments, “ADC12J4000, 12-Bit, 4GSPS ADC With Integrated DDC“, SLAS989D – January
2014 – Revised October 2017., pp. 1-96.
[37] L. Karlsson, “Method, System and Apparatus for Maximizing a Jammer’s Time-on-Target and Power-on-
Target”, United States Patent Application, Publication No. US 2006/0164283 A1, 27. July 2006.
top related